OpenAI Propaganda, Hidden AI Triggers, Copilot’s Santa Fantasy: The AI Argument EP83

Wired Magazine says OpenAI’s “research” is starting to sound like propaganda. Frank says that’s what happens when IPO pressure creeps in — the bad news gets buried unless regulation forces transparency. Justin says Frank’s chasing the wrong villain: regulation won’t save you if politicians still don’t have a plan for the job shock.

Plus: Nick Huber trying to ban AI-written emails. ChatGPT rescuing a missing recording with a bit of command line wizardry. Gemini Flash vs Pro, and whether Gemini 3 Pro is the best “brains per euro”. A nasty “backdoor” idea where harmless training data can hide weird triggers. And Microsoft Copilot getting absolutely mugged by its own Christmas ad.

00:44 Is Nick Huber right to ban AI emails?
03:32 Did ChatGPT just save the podcast?
08:22 Is ChatGPT Images as good as Nanobanana Pro?
10:12 Is Gemini Flash almost as good as Pro?
12:29 Is Gemini 3 Pro the best value for intelligence?
15:34 Could safe training data still hide AI backdoors?
26:31 Is OpenAI research turning into propaganda? 
35:11 Does Microsoft Copilot live up to its ads?

► SUBSCRIBE
Don't forget to subscribe to our channel for more arguments

► LINKS TO CONTENT WE DISCUSSED 

• The new ChatGPT Images is here
• Gemini 3 Flash: frontier intelligence built for speed
• Weird Generalization and Inductive Backdoors: New Ways to Corrupt LLMs
• Company’s Economic Research Is Drifting Into AI Advocacy
• Microsoft’s holiday Copilot ad is wrapped in empty promises

► CONNECT WITH US
For more in-depth discussions, connect Justin and Frank on LinkedIn.
Justin: https://www.linkedin.com/in/justincollery/
Frank: https://www.linkedin.com/in/frankprendergast/